For two decades, the mental model for online visibility has been simple: get links, build authority, rank higher. That model still has some relevance for traditional search rankings — but it has almost nothing to do with how AI systems decide who to recommend.

When an AI assistant answers a question about a business, a place, a product, or a topic, it isn't crawling the web in real time and weighing up which pages have the most backlinks. It's drawing on structured knowledge — records that describe entities, their properties, and how they relate to other entities. A restaurant, a software tool, a historical figure, a local landmark: each of these can exist as a distinct, defined entity in a knowledge base, independent of any specific webpage.

Two different questions

Traditional SEO answers the question: "which page is most authoritative for this search term?" That's a link-and-content question.

AI recommendation answers a different question: "what is this thing, and is it the kind of thing worth mentioning here?" That's an entity question. A page can rank well for a search term while the underlying entity remains completely unknown to the systems that power AI answers.

The shift in one sentence: traditional SEO optimises a page to be found; entity engineering establishes a thing to be known.

Why this matters now

AI Overviews, ChatGPT's browsing features, and Perplexity all lean heavily on structured data sources that long predate generative AI — these sources were built for the Knowledge Graph era, and AI systems inherited them. A website that has never been connected to that infrastructure is operating one layer below where these systems actually look.

This isn't a reason to abandon content or traditional SEO — it's a reason to recognise that there's a separate, largely invisible layer of work that determines whether you're recognised as a "thing" at all, regardless of how well your pages are written or how many links point to them.